The fact that I have made no use of the term analogy in this study, finally, requires a word of comment. The term and the notion have been so widely used to describe relations among surface forms at different stages of a language - that is, to describe diachronic correspondences - that they are best avoided altogether in a theory of change that consistently analyses diachronic correspondences into sequences of innovations. There are deductive innovations which can be described as analogical. But these are accounted for by referring to the synchronic rules that produce them, and not by citing similar surface forms.
- Andersen (2004), a pag.438